Google announced on Tuesday that its flagship AI model Argon is officially open to developers and enterprise customers worldwide. Previously, the model had only been available for testing access to a select group of trusted cyber defense professionals through Google's Fairwind program. This phased release balances security evaluation with commercial promotion, and it also reflects the importance frontier models place on risk control when expanding their openness.
Currently, Argon is available to paid users through the Google AI Studio API, the Google AI Ultra and Pro subscription services in the Gemini App, and the agent development platform Antigravity. Its pricing is 2 dollars per million input tokens and 10 dollars per million output tokens, and the price will double starting in December.
Even after the price increase, its price remains lower than that of OpenAI GPT-6Astra mentioned in the article (10 dollars per million input tokens and 50 dollars per million output tokens). However, according to reports, some Google employees still have doubts about Argon's performance in areas such as coding.
Google DeepMind head Koray Kavukcuoglu said that internal software engineers have already used Argon to deliver high-quality code faster, and the model has also been used to optimize Chrome's user experience. Google hopes to use this to demonstrate the model's value in real engineering tasks. Kavukcuoglu also said that Argon, combining multimodal reasoning, agentic coding, and physical understanding capabilities, is expected to advance the research and development of embodied robotics.
In one demonstration, researchers reconstructed the scene layout and motion trajectories by analyzing videos of humans throwing sandbags, and passed the computational results to a humanoid robot, which then hit the target seven times in a row in a real environment.
Previously, Google's Gemini 3.5 Pro was reportedly delayed due to capability optimization, and the company also saw senior researchers depart. Facing competitors such as OpenAI, Anthropic, and Meta continuously launching new models, Google still needs to prove that Argon not only has potential in engineering and robotics, but can also win over a broader range of users in terms of everyday task experience, performance, and usability.